Originally posted by: Matt1970
We are gonna have a computer in 50 years that is more powerful than all the brains in the world ?
Yep, I'd say it's pretty much guaranteed, we just have to maintain the same pace we have in the last 50 years. The pace is likely to accelerate so you're looking at potentially 30-40 years.
Depends on what you mean by a computer being "powerful". Comparing the processing power of a neural net and an algorithmic processor is difficult at best. For example, the human brain operates at around 100 neuron propagation cycles per second. A processor weighs in at several gigahertz, but can still only carry out a fraction of the tasks the brain can. The CPU is good at well defined, precise calculations such as the operations of formal logic or arithmetic. In this, it can beat the human many times over. A CPU can, however, never be "smart", by definition. A CPU can carry out tasks that someone else (likely a human brain) has already figured out how to do, such as an algorithm for finding prime numbers, calculating the integral of a complicated function, or how to render a number of polygons accurately on a 2D display. A neural network such as our brain, on the other hand, can't do any of those things very well or quickly, but is on the other hand capable of coming up with said algorithms in the first place. The two are intrinsically different, and they excel at different tasks.
Now, could you program a CPU to act and respond exactly as a given neural network would? Certainly, but it would be tremendously difficult and require ridiculous amounts of precessing power (in fact, the easiest way to do it would probably be to "simulate" the neural network in question down to the sub-neuron level, effectively creating an equivalent neural network). But it is possible. It just isn't very efficient or even desirable.
On the other hand, could you make a neural network do what a CPU does? Certainly. In fact, every operation carried out by a modern CPU had first been carried out by a human designer, to check that it accomplishes something desirable. We wouldn't be able to make a cpu unless we knew what it did, and how. But can we do it particularly well, or quickly? Certainly not. A C2D can calculate primes a LOT faster than i ever could. Making a neural network that can accomplish the same things as a CPU isn't efficient or desirable either.
The question of "which is faster, the new XXXz0rz Super Deluxe Ultra Plus Extra Gigacore 100 000 GHz processor, or all the human brains in the world combined?" is really a meaningless question. It depends entirely on what you want to accomplish.